DIVISION OF THE HUMANITIES AND SOCIAL SCIENCES CALIFORNIA INSTITUTE OF TECHNOLOGY PASADENA, CALIFORNIA 91125 WHEN IS THE CORE EQUIVALENCE THEOREM VALID? Charalambos D. Aliprantis California Institute of Technology Indiana University - Purdue Univeristy at Indianapolis Owen Burkinshaw Indiana University - Purdue University at Indianapolis SOCIAL SCIENCE WORKING PAPER 707 September 1989 WHEN IS THE CORE EQUIVALENCE THEOREM VALID? Charalambos D. Aliprantis and Owen Burkinshaw ABSTRACT In 1983 L. E. Jones exhibited a surprising example of a weakly Pareto optimal al­ location in a two consumer pure exchange economy that failed to be supported by prices. In this example the price space is not a vector lattice (Riesz space). Inspired by Jones' example, A. Mas-Colell and S. F. Richard proved that this pathological phenomenon cannot happen when the price space is a vector lattice. In particu­ lar, they established that (under certain conditions) in a pure exchange economy the lattice structure of the price space is sufficient to guarantee the supportability of weakly Pareto optimal allocations by prices-i.e., they showed that the second welfare theorem holds true in an exchange economy whose price space is a vector lattice. In addition, C. D. Aliprantis, D. J. Brown and 0. Burkinshaw have shown that when the price space of an exchange economy is a certain vector lattice, the Debreu-Scarf core equivalence theorem holds true, i.e., the sets of Walrasian equilib­ ria and Edgeworth equilibria coincide. (An Edgeworth equilibrium is an allocation that belongs to the core of every replica economy of the original economy.) In other words, the lattice structure of the price space is a sufficient condition for avoiding the pathological situation occuring in Jones' example. This work shows that the lattice structure of the price space is also a necessary condition. That is, "optimum" allocations in an exchange economy are supported by prices (if and) only if the price space is a vector lattice. Specifically, the following converse-type result of the Debreu-Scarf core equivalence theorem is established: If in a pure exchange economy every Edgeworth equilibrium is supported by prices, then the price space is neceBSarily a vector lattice. WHEN IS THE CORE EQUIVALENCE THEOREM VALID? Charalambos D. Aliprantis and Owen Burkinshaw* An answer to the question of the title is provided. It is shown that under some conditions the core equivalence theorem holds true (ifand ) only if the price space is a vector lattice. 1. INTRODUCTION Many of the current models of economies with infinitely many commodities require that the commodity space is a Riesz space and that the commodity-price dual pair is a Riesz dual system. Recently, there has been interest in how much these requirements can be relaxed. It is our purpose to examine how the basic welfare theorems relate to the lattice structure of the commodity and price spaces. We consider economies where the commodity space E is a vector lattice with a locally convex topology and the price space is the topological dual E' with a generating positive cone. The main result of this paper asserts that if the core equivalence theorem holds true for such economies, then the price space must be a vector lattice. This shows that the lattice structure of the commodity-price duality is essential for the validity of the fundamental theorems of welfare economics. In this paper, we shall employ the mathematics of Riesz spaces. We shall follow the notation and terminology of [5] . Let us briefly mention afew things about Riesz spaces. A Riesz space (or a vector lattice) is a partially ordered vector space which is also a lattice-in the sense that for each pair of vectors x and y the supremum (least upper bound) and the infimum (greatest lower bound) exist; using standard lattice terminology we shall write x Vy= sup{x,y} and x I\ y= inf{x,y}. As usual, the positive cone of a Riesz space E will be denoted by E+, i.e., E+= {x EE: x 2: 0}-the symbol lR will denote the set of real numbers. For an arbitrary element x of a Riesz space, its positive part, its negative part, and its absolute value are defined by the formulas x+=xVO, x-=(-x)VO, and lxl=xV(-x) . * Research of both authors was supported in part by a Chrysler Corporation grant to IUPUI. 2 A vector subspace F of a Riesz space E is said to be a Riesz subspace whenever for each x, y E F the supremum xV y (taken in E) belongs to F. A (non-empty) subset A of a Riesz space is said to be a solid set whenever Jxl :":'. JyJ and y E A imply x E A. A solid vector subspace of a Riesz space is referred to as an ideal. An ideal is always a Riesz subspace but the converse is not true. A linear functional f:---> iR on a Riesz space E is said to be order bounded whenever it carries order bounded subsets of E onto bounded subsets of iR; every set of the form [x, y] = {z E E: x :":'. z :":'. y} is known as an order interval and subsets of order intervals are called order bounded sets. Every positive linear functional f, i.e., any linear functional f that satisfies f(x) 2: 0 for each x 2: 0, is necessarily order bounded. The set of allorder bounded linear functionals on a Riesz space is a vector space which is also a partially ordered vector space under the ordering f 2: g whenever f( x) 2: g( x) for all x E E+. The partially ordered vector space of all order bounded linear functionals on a Riesz space E is called the order dual of E and is denoted by E-. It turns out that the the order dual E' of a Riesz space E is also a Riesz space whose lattice operations are given by fVg(x)=sup{f(y)+g(z): y, zEE+ and y+z=x} and f/\g(x)=inf{f(y)+g(z): y, z EE+ and y+z=x} for aii f, g E E- and all x E E+. Definition 1.1. A Riesz dual system (E, E') is a dual system such that 1) E is a Riesz space; 2) E' is an ideal of the order dual E- separating the points of E; and 3) the duality function ( · , ·) is the natural one, i. e., (x,x')= x'(x) holds for all x E E and all x' E E'. Most examples of dual systems employed in economics are Riesz dual systems. Here are a few examples of Riesz dual systems. a) (Lp(µ),Lq(µ)), l<p,q<oo; �+�=1; b) (Cp,Cq), l:"'.'.p,q:"'.'.oo; �+�=1; c) (L=(µ),L1(µ)) and (L1(µ),L=(µ)), µ a a-finitemeasure; and d) (C(ll), ca(ll)), l1 a Hausdorff compact topological space. The dual system (ca(ll), C(ll)) is not (in general) a Riesz dual system. A symmet­ ric Riesz dual system is a Riesz dual system (E, E') such that each order interval of E is weakly compact (i.e., a(E, E')-compact). The term "symmetric Riesz dual system" is justified from the fact that in a Riesz dual system (E, E') the order in­ tervals of E are wealdy compact if and only if E is (under its natural embedding) an ideal of (E'r, and so (E ' ,E) is also a Riesz dual system. 3 A locally solid Riesz space (E, r) is a Riesz space E equipped with a locally solid topology r (a linear topology r is said to be locally solid whenever it has a base at zero consisting of solid neighborhoods). The topological dual E' of a locally solid Riesz space (E, r) is always an ideal of the order dual E', and hence E' is always a Riesz subspace of the order dual. A locally convex-solid Riesz space (E, r) is a Riesz space E equipped with a Hausdorff linear topology r that has a base at zero consisting of solid and convex neighborhoods-the topology r is referred to as a locally convex-solid topology. It turns out that a linear topology on a Riesz space space is locally convex-solid if and only if it is generated by a family of lattice seminorms. (A seminorm q on a Riesz space is said to be a lattice seminorm whenever lxl :<::: IYI implies q(x) :<::: q(y).) If (E, r) is a locally convex-solid Riesz space and E' denotes its topological dual, then (E, E') is a Riesz dual system-as a matter of fact, every Riesz dual system (E, E') can be obtained in this manner, i.e., if (E, E') is a Riesz dual system, then there exists a Hausdorff locally convex-solid topology r on E consistent with the duality (E, E') such that E' is the topological dual of (E, r). For a given Riesz dual system (E, E') there are two distinguishable consistent locally convex-solid topologies on E. They are the absolute weak and absolute Mackey topologies. The absolute weak topology l<Tl(E, E') is the locally convex-solid topology on E of uniform convergence on the order intervals of E' and is generated by the family of lattice seminorms {qx'' x' EE'}, where qx'(x) = lx'l(lxl) for all x E E and all x' E E'. The absolute Mackey topology lrl(E, E') is the locally convex-solid topology on E of uniform convergence on the convex, solid, and <T( E', E)-compact subsets of E'.